[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [Xen-devel] (partial) Spectre v2 mitigation without on Skylake IBRS
On 26/02/18 11:49, Jan Beulich wrote: >>>> On 26.02.18 at 11:18, <jgross@xxxxxxxx> wrote: >> On 26/02/18 10:44, Jan Beulich wrote: >>> if running PV Linux on older Xen (4.5 and earlier) is relevant, it may be >>> necessary to use a mechanism other than IBRS to mitigate Spectre v2 >>> on Skylake. That is because the new MSR value can't be migrated >>> prior to migration v2. Of course one option would be to retrofit some >>> mechanism into newer Xen versions that makes them accept whatever >>> extension to e.g. struct hvm_hw_cpu one might want to invent for >>> the older Xen versions. But that doesn't seem very desirable. >> >> Can you please elaborate a little bit more what the real problem is? >> >> I _think_ you are referring to the problem that a pv kernel would want >> to use IBRS for mitigation of Spectre V2 and after a migration that >> setting would be lost. > > "Lost" is the wrong term imo: A hypervisor that's been patched for > Spectre v2 (and that's a prereq anyway, because we want the > kernel to use IBPB, which utilizes an MSR that doesn't need > migrating) should at least do _something_ with the MSR (when it's > non-zero). The most natural thing (imo) is to make those older > hypervisors support XEN_DOMCTL_{get,set}_vcpu_msrs. That in > turn calls for the tool stack to gain the check that Andrew had > added to libxc in db24f7f012 ("libxc: use an explicit check for PV > MSRs in xc_domain_save() "), causing migration to fail when the > MSR is non-zero on any of the guest's vCPU-s. > >> If this is the case I believe the easiest solution would be to let the >> kernel set the MSR again after leaving suspended state. suspend/resume >> require hooks in pv kernels after all. > > Hmm, this could be leveraged irrespective of what I've written > above - the kernel could then also clear the MSR during suspend, > thus allowing the check in libxc to pass. Something like the attached patch? Juergen Attachment:
0001-x86-xen-zero-MSR_IA32_SPEC_CTRL-before-suspend.patch _______________________________________________ Xen-devel mailing list Xen-devel@xxxxxxxxxxxxxxxxxxxx https://lists.xenproject.org/mailman/listinfo/xen-devel
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |